Physical Intelligence, the San Francisco robotics AI startup building general-purpose models for robot control, is reportedly in discussions to raise approximately $1 billion in new funding at a valuation above $11 billion, according to fundsforNGOs News. The round would nearly double the company’s previous $5.6 billion valuation from earlier this year.

Investors and Scale

The funding round has reportedly attracted interest from Founders Fund and Lightspeed Venture Partners, along with existing backers Thrive Capital and Lux Capital. Discussions remain ongoing and final terms could change, according to the report. Physical Intelligence has previously raised more than $1 billion in total funding and employs approximately 80 people.

Founded by Karol Hausman, Sergey Levine, and Chelsea Finn, the company has taken a research-first approach, investing in foundational AI for robotics rather than rushing to commercialize. The bet: build models capable of controlling any robot to do any task, then let the applications follow.

Technical Progress

That bet has recent evidence behind it. In April 2026, Physical Intelligence released π0.7, a steerable robotic foundation model that the company says exhibits “a step-change in generalization.” Unlike prior vision-language-action models that required fine-tuning for each specific skill, π0.7 demonstrates compositional generalization, recombining skills from its training data to solve tasks it has never seen before.

The model can follow new language commands, use unfamiliar kitchen appliances, and even enable a new robot platform to fold laundry without laundry-specific training data for that robot. Physical Intelligence describes this as “the first signs of compositional generalization” in robotic foundation models, a capability analogous to how large language models can combine concepts (translating English to French while formatting output as JSON) but applied to physical manipulation.

Commercial Deployment

The company has also begun testing its models in production environments. In February 2026, Physical Intelligence published a partner deployment report showing its π0.6 model running in a commercial laundromat in San Francisco through a partnership with Weave Robotics. The system folds laundry autonomously, handling deformable, variable garments in a live customer environment.

Physical Intelligence frames these deployments as stress tests for its general-purpose models rather than finished products. The company’s stated vision is to create a “physical intelligence layer” comparable to foundation model APIs in software: a ready-made stack that any robotics developer can build on, rather than constructing custom controllers, data pipelines, and models from scratch for each application.

Capital Concentration in Embodied AI

If completed, the round would place Physical Intelligence among the most heavily capitalized AI robotics startups globally. The timing aligns with a broader capital surge into AI: Crunchbase reported that AI startups attracted $242 billion in venture funding in Q1 2026 alone, representing 80% of all global VC capital deployed. Within that wave, embodied AI and robotics have emerged as a distinct capital-attracting category, with investors betting that agentic reasoning applied to physical tasks represents the next frontier beyond software-only agent systems.

The round remains in discussions. Physical Intelligence has not publicly commented on the funding.